My first LoRA training. I wanted to create a lora of Yaotome Urushi, from Soredemo ayumu wa yosetekuru anime.
Used 23 images, 512x512 and 100 steps per image with Kohya UI.
For being my first LoRA training, i think it has good results. I'll doing it better in future anyways.
Basic structure:
<lora:yaotome_urushi1:0.8>, yaotome_urushi, school_uniform, brown_eyes ..."
0.8 or lower, Lower CFG Scale is better
Used 7th_anime_v3_C model for training (https://huggingface.co/syaimu/7th_Layer)
PD: Sometimes, eyes might not be brown, i think because for the short training.